ワンクリックで
am-command-skillshare-recommend
Analyze codebase with repomix, search skillshare for relevant skills, and recommend top 5
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
メニュー
Analyze codebase with repomix, search skillshare for relevant skills, and recommend top 5
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
Reference codebase for Pi Extensions. Use this skill when you need to understand the structure, implementation patterns, or code details of the Pi Extensions project.
Initialize non-obvious modular rules for the codebase
Plan and add as task
SOC 職業分類に基づく
| name | am-command-skillshare-recommend |
| description | Analyze codebase with repomix, search skillshare for relevant skills, and recommend top 5 |
| x-agentsmesh-kind | command |
| x-agentsmesh-name | skillshare-recommend |
| x-agentsmesh-allowed-tools | ["Read","Grep","Glob","Bash"] |
Generate a skill profile of this repository using repomix, then search the skillshare registry for relevant community skills and recommend the best matches.
Focus on domain-relevant skills: ones that match what this project is (e.g., skill systems, extension patterns, package management, skill authoring, configuration management). Deprioritize generic tooling, CI, and AI/LLM skills unless they directly match a unique aspect of the project.
Determine repo name: Extract the repository name from git remote -v (e.g., origin git@github.com:user/repo.git → repo). Use that as <repo name>.
Run repomix: Execute:
npx -y repomix --skill-generate --no-dot-ignore --skill-output .skillshare/skills/<repo name> --force
Analyze generated files: Read the generated files in .skillshare/skills/<repo name>/ — examine the skill profile, detected patterns, languages, frameworks, and conventions captured by repomix.
Extract tech stack signals: From the repomix-generated skill profile, extract every distinct:
Populate up to 3 queries and run multi-search aggregation: From the repomix tech-stack, craft at most 3 queries — they should target domain-relevant skills (e.g., skill systems, extension patterns, package management, skill authoring, configuration management). Avoid queries about tooling, CI, testing, or generic AI/LLM topics unless they match a unique project need. Execute the script below.
# ── Populate with 1-3 domain-targeted queries ──
QUERIES=(
"skill system"
"extension pattern"
"configuration management"
)
# ── Automated aggregation script ──
RESULTS_DIR=$(mktemp -d)
ALL="$RESULTS_DIR/_all.json"
echo '[]' > "$ALL"
echo "=== Running ${#QUERIES[@]} targeted searches ==="
for q in "${QUERIES[@]}"; do
safe=$(echo "$q" | tr -s ' /' '_')
out="$RESULTS_DIR/$safe.json"
echo " Searching: $q"
skillshare search "$q" --json --limit 10 > "$out" 2>/dev/null
if [[ $? -ne 0 || ! -s "$out" ]]; then
echo " ⚠ Empty"
continue
fi
is_valid=$(jq 'if type == "array" and length > 0 then true else false end' "$out" 2>/dev/null)
if [[ "$is_valid" != "true" ]]; then
echo " ⚠ No results"
continue
fi
# Tag each result with its source query then merge
jq --arg q "$q" '[.[] | . + {_source: $q}]' "$out" > "${out}.tagged" 2>/dev/null
jq -s 'add' "$ALL" "${out}.tagged" > "${ALL}.tmp" 2>/dev/null && mv "${ALL}.tmp" "$ALL"
done
echo ""
echo "=== Aggregating and deduplicating ==="
AGGREGATED="$RESULTS_DIR/aggregated.json"
# Deduplicate by Name, sort by Stars descending, track matched queries
jq '
group_by(.Name)
| map(
first as $orig
| {
name: $orig.Name,
description: $orig.Description,
stars: ($orig.Stars // 0),
owner: $orig.Owner,
source: $orig.Source,
repo: $orig.Repo,
tags: $orig.Tags // [],
matched_by: [.[] | ._source] | unique
}
)
| sort_by(-.stars)
| to_entries
| map(.key += 1 | { rank: .key, name: .value.name, description: .value.description, stars: .value.stars, owner: .value.owner, source: .value.source, repo: .value.repo, tags: .value.tags, matched_by: .value.matched_by })
' "$ALL" > "$AGGREGATED" 2>/dev/null
if [[ ! -s "$AGGREGATED" ]]; then
echo "No valid results could be aggregated."
cat "$ALL" 2>/dev/null | head -c 2000 || true
rm -rf "$RESULTS_DIR"
exit 1
fi
echo ""
echo "╔══════════════════════════════════════════════════════════╗"
echo "║ SKILLSHARE RECOMMENDATIONS ║"
echo "╚══════════════════════════════════════════════════════════╝"
jq -r '
.[] |
"\(.rank). \(.name)
⭐ \(.stars) stars | Owner: \(.owner // "?")
\(.description // "(no description)")
Matched by: \(if (.matched_by | length) > 0 then (.matched_by | join(", ")) else "direct search" end)
"' "$AGGREGATED" 2>&1 | head -120
cp "$AGGREGATED" ".skillshare/skills/<repo name>/aggregated-results.json"
echo ""
echo "Full results saved to: .skillshare/skills/<repo name>/aggregated-results.json"
rm -rf "$RESULTS_DIR"
Review and recommend top 5: Read the leaderboard from the script output. Pick the 5 most relevant skills based on:
For each of the 5 recommendations, explain: